Memory copy loop optimization

[k06a]

This loop could be replaced with CALLDATACOPY():

clones-with-immutable-args/src/ClonesWithImmutableArgs.sol

Lines 126 to 134 in 5950723

	for (; counter >= 32; counter -= 32) {
	// solhint-disable-next-line no-inline-assembly
	assembly {
	mstore(copyPtr, mload(dataPtr))
	}
	copyPtr += 32;
	dataPtr += 32;
	}

[hrkrshnn]

Alternatively, if one really wants to keep the data in memory, one can use the identity precompile to copy from memory to memory. It should be cheaper if that loop runs more than 2 times (please check).

[k06a]

Yep, I see method is internal and argument is not calldata, then really consider using “identity” precompile for copying

[BlinkyStitt]

If that loop is replaced, copyPtr and dataPtr both need to be increased still, right?

And I think counter needs to be decreased, too. Otherwise mask will be wrong

[z0r0z]

this looks great. any thoughts on integrating @wighawag ?

[0xca11]

Implemented this optimization using the identity precompile on this fork in case anyone's interested: RollaProject@3cef7ae

@z0r0z @wighawag Please let me know if a PR with this change would be welcome :)

[BlinkyStitt]

How much did deploy and call costs change with the identity precompile?

…

On May 23, 2022, at 4:16 AM, 0xca11.eth ***@***.***> wrote:  Implemented this optimization using the identity precompile on this fork in case anyone's interested: ***@***.*** @z0r0z @wighawag Please let me know if a PR with this change would be welcome :) — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.